摘要
正确识别超声图像中的甲状旁腺结节对甲状旁腺功能亢进的诊断治疗非常重要.由于病人个体的差异性和超声图像的复杂性,采用图像的形态特征和纹理特征识别甲状旁腺结节准确率低.本文提出利用包膜以及结节与甲状腺相对位置的先验知识特征描述方法,并将其与形态、纹理特征融合,采用支持向量数据描述(Support Vector Data Description,SVDD)识别甲状旁腺结节.实验结果表明,先验知识特征可以准确描述甲状旁腺结节的特征,融合先验知识特征比仅利用形态特征和纹理特征具有更高的识别准确率.
It is very important to recognize parathyroid nodules correctly in ultrasound images for the treatment of hyperparathyroidism.Due to individual differences of patients and complexity of ultrasound images,parathyroid nodules can’t be recognized accurately by only using morphological features and texture features.In this paper,a prior knowledge feature description method is proposed on account of the characteristic of envelope and the relative location between the nodule and the thyroid.SVDD is applied to recognize parathyroid nodules based on the fusion features of prior knowledge features,morphological features and texture features.The experimental results show that the prior knowledge features can describe the characteristics of parathyroid nodules well,and the accuracy by using the fusion features which combined prior knowledge features is higher than that of only using morphological features and texture features for the recognition of parathyroid nodules.
作者
毛林
赵利强
于明安
魏莹
王颖
MAO Lin;ZHAO Li-qiang;YU Ming-an;WEI Ying;WANG Ying(College of Information Science and Technology,Beijing University of Chemical Technology,Beijing 100029,China;Interventional Ultrasound Medicine,China-Japan Friendship Hospital,Beijing 100029,China;Institute of Microelectronics of Chinese Academy of Sciences,Beijing 100029,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《电子学报》
EI
CAS
CSCD
北大核心
2021年第5期944-952,共9页
Acta Electronica Sinica
基金
北京化工大学-中日友好医院生物医学转化工程联合基金(No.PYBZ1804)。
